Inside AI - November 19th, 2019

Inside AI (Nov 19th, 2019)

Megvii IPO / Cerebras AI computer / DeepMind AI on Play Store recommendations

Subscribe | View in browser

1. China-based Megvii Technology Ltd plans to seek listing approval this week on a Hong Kong IPO and is seeking to raise over $500 million, according to a report from Reuters. The AI company, which developed the facial recognition tech Face++, filed its IPO prospectus with the Hong Kong Stock Exchange this past summer. Shortly after, it was blacklisted by the U.S. government along with seven other Chinese AI companies over their alleged involvement in human rights violations of Muslim minorities in China's Xinjiang Uighur Autonomous Region. The company still plans to move forward on an IPO, which would make it the first Chinese AI firm to go public. A $750 million fundraising earlier this year valued the company at just over $4 billion. - REUTERS

2. Cerebras unveiled today what it claims is the industry’s fastest AI computer. The CS-1 can replace hundreds of racks of GPU-based computers using hundreds of kilowatts of power, according to the company, but measures only 26 inches tall and consumes about 20 kilowatts itself. Andrew Feldman, Cerebras CEO, says that the system dramatically cuts down on training time and, in the case of deep neural networks, can achieve single image classification in microseconds. The software side of the system allows users to write machine learning models using standard frameworks like Pytorch and Tensorflow. Already, Argonne National Labs, which plans to house a future exascale supercomputer, said it has deployed a CS-1 to accelerate neural networks in cancer studies and study traumatic brain injuries, among other uses. - TECHCRUNCH

3. An AI dermatology app that identifies skin conditions doesn't appear to work effectively on people with dark skin, according to a new study from researchers in Uganda and Sweden. Users can upload photos on the AI-based Skin Image Search, which suggests the three most likely skin diseases with an accuracy of about 80 percent for the latest algorithm version. However, researchers tested an earlier version of the algorithm on adults in Uganda and found that it failed to make correct suggestions on 102 of the 123 photos that were uploaded. The convolutional neural network was trained on more than 300,000 images of skin diseases, although black skin makes up only about 5 to 10 percent of the database. - PHYSICS WORLD

4. In a blog post today, DeepMind researchers explain how they use AI and machine learning model architectures to improve recommendations for apps in Google's Play Store. The company says the new system was deployed earlier this year and makes recommendations that are more personalized than they were previously. The system contains three main models: a candidate generator, a reranker, and an AI model "to optimize for multiple objectives." As the blog explains, the candidate generator "is a deep retrieval model that can analyze more than a million apps and retrieve the most suitable ones." A reranker, or user preference model, "predicts the user's preferences along multiple dimensions," and the predictions are fed into a multi-objective optimization model "whose solution gives the most suitable candidates to the user." - VENTURE BEAT

5. Alain Bénichou, CEO of IBM’s Greater China business, says that only about 14 percent of companies in China are actively using AI. Bénichou made the remarks on Tuesday during CNBC’s East Tech West conference in Guangzhou, China. While China is adopting AI at a faster rate than the world's average, there is still "a long way to go,” Bénichou said, adding that there is a “problem with skills" and talent at Chinese companies, which he claimed are struggling to go beyond the proof of concept stage and are being held back by a lack of trust in AI enterprise systems, which aren't always transparent. “There is a craving for AI,” said Bénichou, “The problem is how to get there." - CNBC

6. TD Ameritrade Holding Corp. has rolled out an AI-powered tool that personalizes users' investment strategies, according to The Wall Street Journal. The brokerage firm wants "to be more of a Netflix-style or Spotify-style personalization engine for digital investing,” CIO Vijay Sankaran said. A team of eight AI experts at the company developed the tool, which sends personalized emails to investors based on their profiles and portfolios. Sankaran said the company prefers to develop AI tools in house, rather than hiring an outside vendor, so that they can customize them for the financial industry. The team is now working on an AI engine dubbed “voice of the customer analytics,” which is being trained on email, voice, and chat transcripts between customers and agents in order to classify the intent of callers, Sankaran said. - WSJ

7. Japanese university researchers partnered with IBM to identify a new geoglyph among the Nazca Lines in Peru. The geoglyphs of humanoids, geometric shapes, and animals date from 500 BC to 500 AD and were drawn by pre-Incan people. To identify more glyphs, a team from Yamagata University in Japan used LiDAR, images from satellites and drones, and geographical surveys, then fed the data into the IBM PAIRS Geoscope, a cloud-based AI system, which evaluated it in a short time period. The results helped researchers uncover a new 15-foot-long geoglyph of a humanoid figure, as well as another 500 candidate sites for new Nazca Lines. - ANCIENT ORIGINS

8. Scientists at MIT’s Computer Science and Artificial Intelligence Laboratory developed an AI system for self-driving cars that can classify the social personalities of human drivers around it, potentially improving its predictions about dangers. The scientists used a social psychology tool known as social value orientation, which ranks how selfish or cooperative a human driver is on the road. Their algorithm estimates the orientation of other drivers to help the self-driving vehicle maneuver better. In tests involving merging lanes and unprotected left turns, the algorithm could better predict the behavior of other cars by a factor of 25 percent. - ZDNET

9. Salesforce plans to utilize technology from Amazon Web Services to improve its customer service apps, the companies announced on Tuesday. The AI system from Amazon's cloud computing unit can translate a customer's spoken words into text as well as different languages, and analyze those texts to determine a customer's feelings, like anger. Salesforce will use its own software - the intelligence layer known as Einstein - to make suggestions for how to respond to the customer. Salesforce's partnership with Amazon Web Services dates back to 2016, when the two signed a $400 million infrastructure services agreement. - REUTERS

10. Forrester vice president and principal consultant Huard Smith developed a list of professions that he believes will be largely replaced by automation by 2030. These include what Smith projects are 73 percent of all cubicle-related jobs, or about 20 million jobs, and 38 percent of location-based jobs, such as grocery store clerks, which equates to nearly 30 million positions. By 2030, Smith believes that AI will eliminate nearly 30 percent of all jobs in the U.S. while creating the equivalent of only 13 percent. "It will be a difficult 10 years and beyond, and the world doesn’t just stop by 2030, so buckle up," he said. - FORTUNE

Written and curated by Beth Duckett in Orange County. Beth is a former reporter for The Arizona Republic who has written for USA Today, Get Out magazine and other publications. Follow her tweets about breaking news and other topics in southern California here.

Editor: Kim Lyons (Pittsburgh-based journalist and managing editor at Inside).

Copyright © 2020, All rights reserved.

Our mailing address is:
767 Bryant St. #203
San Francisco, CA 94107

Did someone forward this email to you? Head over to to get your very own free subscription!

You received this email because you subscribed to Inside AI. Click here to unsubscribe from Inside AI list or manage your subscriptions.

Subscribe to Inside AI